Reading Honda datalogs: KPro, FlashPro and FuelTech
Published
A Honda log records sensor readings and ECU commands. The channel name tells you what was recorded; identifying the cause takes more work. This page covers port-injected K20 and K24 logs in KManager, FlashProManager and FuelTech. Validate the file, mark the control state, compare values with their targets and test one cause at a time. Targets and sensor limits come from the actual calibration and hardware.
- Keep the native log and the CSV. The CSV is an export, not the master record.
- A missing channel is missing, not zero. KManager’s documented CSV does not include every sensor the software displays.
- Read the control state before the number: startup, closed loop, fuel cut, VTEC and the cam transition.
- Compare lambda with
targetlambda, not with a number copied from another engine. - KManager’s
mapis in bar and the factory Honda sensor reads absolute pressure. Subtract the local atmosphere before calling it boost. - One run covers a narrow path through rpm, load and cam angle. Do not tune every cell from one trace.
Full Custom: $249 $199 launch special. Most files in 3–5 business days after complete inputs; the outside limit is 7 business days.
Port-injected K20 and K24 only. The turbo Civic is the separate turbo section on the same Honda & Acura hub. Hondata or FuelTech hardware and credits are not included. How files are handled.
1. The boundary: port-injected K20 and K24
This page covers a port-injected K20 or K24. The software can be KManager on KPro, FlashProManager on FlashPro, or FTManager on a FuelTech FT450, FT550 or FT600. What matters is that the calibration matches the engine, the sensors and the throttle.
K20C, L15, B/D-series, s300 and direct-injection K24W logs are outside this guide. Those use different fuel systems or controllers. Matching column names do not establish matching measurements.
There is no universal K idle speed, mixture, pressure, cam angle or knock allowance. Use the calibration, injector data and sensor scaling for this build.
An ECU log records what the ECU saw and what it commanded. It does not record cylinder pressure, individual exhaust temperature, per-injector fuel mass or the mechanical clearance of a cam. A finding has to stay inside the measurement that supports it.
2. What the file holds, and what it does not
KManager saves a native .kdl log and exports the loaded log to CSV. FlashProManager saves a native .fpdl log. FTManager saves PowerFT logs as .ftml and can export CSV. CSV does not necessarily carry the calibration or every status channel. Keep an untouched export; sorting or spreadsheet conversion can change row order and number formatting.
- The KManager help page lists the documented CSV fields.
timeis in seconds,mapis in bar, andclvandtpsare in percent.camangleandtargetcamangleare in degrees,injectoris in milliseconds, andignitionis in degrees before top dead center.taandtware in degrees Celsius.lambdaandtargetlambdaare lambda, andshortermandlongtermare in percent. The list also includesknockcount, batteryvoltagein volts andeldin amps. - The documented list omits VTS, fuel pressure, injector duty, K.Retard and K.Control. That is not proof those channels are absent from the ECU or every current export. Inspect the actual header: exported fields and spellings vary by KManager build.
- FlashPro sensors depend on the ECU. The help matrix lists VTC Actual, VTC Cmd, VTS and knock and loop-status channels for the Civic Si ’06 entry. Check the actual calibration and export; the sensor matrix is not a CSV schema or a promise that every supported ECU exposes those channels.
- FuelTech PowerFT logs are built from the channels configured in the map. A fuel-pressure trace needs a real, characterized pressure sensor assigned to an input. An O2 trace needs a wideband controller and a valid association. If the channel is disabled or absent, it stays missing. It is not zero pressure, zero duty or zero knock.
| Concept | KManager / KPro | FlashProManager | FuelTech PowerFT | What to check |
|---|---|---|---|---|
| Load | map in bar, clv in percent | MAP and calculated load | MAP and TPS from the configured sensor | Absolute or gauge reference, then the active fuel model |
| Mixture | lambda and targetlambda | Lambda and Target Lambda | O2 actual and target | Sensor validity, scale, transport delay and closed-loop state |
| Fuel correction | shorterm and longterm | Short Term Trim and Long Term Trim | O2 closed-loop correction | Sign convention and whether the loop is enabled |
| Delivery | injector in milliseconds | INJ and duty where supported | Bank or cylinder pulse and duty | Event mode, dead time and staged banks |
| Cam | camangle and targetcamangle | VTC Actual and VTC Cmd | Intake phase target and actual if configured | Crank degrees, sensor zero and mechanical travel |
| Lobe state | VTS where the log carries it | VTS where the ECU supports it | Separate output or window state | A command is not proof the rocker moved |
| Knock | knockcount; other knock values may not export | Knock Level, Retard and Control where supported | External knock monitoring if fitted; verify whether its signal is recorded | Strategy, sensitivity, fuel and operating point |
| Spark | ignition in degrees BTDC | IGN in degrees | Ignition timing in degrees | What the command includes, and what it does not prove |
To compare logs, record chassis, engine, ECU code, calibration revision, software build and sensor list. Add injector part number and rated pressure, MAP sensor, wideband controller, fuel blend and the exact change under test. Differences without that context can come from mismatched files rather than the change you intended.
Swipe or use the arrow keys to read every label.
3. Read the control state before the number
Mark operating states before making corrections. The same spark value can be an idle-control action or a load-table result. A lambda reading may describe steady combustion, fuel cut or a transition. Each needs a different comparison.
| State | What the file can show | What it does not prove |
|---|---|---|
| Crank and post-start | Low or zero cam actual, richer command, high idle | A cam fault, a mixture fault or a bad sensor |
| Warm idle | Spark moving as an idle actuator, lambda near target, small trim movement | A universal idle rpm, MAP or spark value |
| Closed loop | Trim and O2 correction moving around the target | A correct base table when the trim is doing the work |
| Fuel cut on overrun | Injector pulse at zero or near zero, lambda reading lean | A lean fuel-table cell |
| VTEC command | An output bit or a configured window state | That the rocker, oil pressure and lobe changed |
| VTC transition | Actual phase moving toward the command | Fault when it lags by a fraction of a second |
| Steady load | A repeatable rpm, MAP and cam region | Coverage of the whole table |
KManager's CAM channel measures intake cam position in degrees of crankshaft rotation from a base position. The help states that the intake is locked in the zero, fully retarded position for the first 10 seconds after startup. A cold-start log that shows zero cam is therefore doing what the calibration asks. It is not evidence of a failed phaser.
- KManager documents 0.1 to 0.5 seconds of possible lag behind
CAMCMD. At 6,000 rpm that is 10 to 50 crankshaft revolutions. Compare movement and settled error before declaring a tracking fault. - Closed loop needs the same care. KManager's documented CSV has no closed-loop status column. FlashPro's sensor list has Closed Loop Status only where the ECU supports it. A PowerFT log shows the O2 correction channel when it is configured. If the state is absent, do not claim the loop was open or closed. Read the calibration's enable condition and the behavior of the trim, and label the limit in the note.
- A VTEC command is a command. It is not proof that oil pressure moved the rocker or that the high cam is carrying the cylinder. If the log has no VTS or equivalent output state, a step in rpm or sound is not a substitute. Treat the lobe change as unproven and say so.
4. Units, references and the traps that survive
Check units and pressure reference before reading a value. KManager exports map in bar, and the factory sensor measures absolute pressure. Standard sea-level atmosphere is 1.013 bar to three decimal places; local atmosphere varies with elevation and weather.
- A log at 1.80 bar absolute is 0.787 bar above standard sea level. Multiply by 14.5038 and that is 11.4 psi of boost, not 26.1 psi. The 26.1 figure is what you get if you multiply 1.80 bar by 14.5038 and forget to subtract the atmosphere. It changes the load region, the fuel demand and the ignition discussion.
- Temperature has the same failure. KManager's
taandtware in degrees Celsius. A coolant reading of 87.8 degrees Celsius is 190 degrees Fahrenheit. A gauge or a note that says 190 degrees Fahrenheit has to be converted before it is compared withtw. Reading 190 degrees Fahrenheit as 190 degrees Celsius overstates the temperature by 102.2 degrees Celsius, enough to call a healthy engine overheated. - Speed has a smaller but real trap. A 60 mph row is 96.6 km/h. A km/h log compared with an mph note moves a gear inference or a road-speed comparison by more than a third.
| Value in the file | What it is | Correct reading | The trap |
|---|---|---|---|
map 1.80 bar | Absolute manifold pressure on a factory Honda MAP sensor | 0.787 bar above standard sea level, or 11.4 psi boost | Calling 1.80 bar gauge pressure and reporting 26.1 psi |
tw 87.8 | Coolant temperature in degrees Celsius in the documented header | 87.8 °C is 190 °F | Reading 190 as Celsius, or 87.8 as Fahrenheit, and calling the engine hot or cold |
| 60 mph | Road speed in miles per hour | 96.6 km/h | Comparing a mph log with a km/h log as if the numbers were the same unit |
| Lambda 0.98 against target 0.90 | Measured mixture against the commanded target in a steady open-loop segment | 0.98 divided by 0.90 is 1.0889, or 8.89% more fuel required to reach the target at unchanged airflow | Using a flat 10% per AFR point and calling it exact |
injector 16 ms at 6,000 rpm | Pulse width on a four-stroke with one injection per cylinder per cycle | 16 times 6,000 divided by 1,200 is 80% duty | Reading pulse width without rpm and calling it duty |
injector 16 ms at 7,000 rpm | The same pulse at a higher engine speed | 93.3% duty with the same event mode | Assuming a pulse that fit at 6,000 rpm still has headroom at 7,000 rpm |
The injector arithmetic is a check, not a replacement for a duty channel. It assumes one injection event per cylinder per four-stroke cycle. With multiple events, sum their pulse widths per cycle; with staged banks, calculate each injector bank separately. Cylinder count does not change this per-injector calculation.
KManager's help defines 100% duty as an injector that never closed and cannot deliver more fuel. The same help suggests planning an injector or fuel-system upgrade once duty reaches 80–85%. That is a KManager planning threshold, not a universal limit for every injector or every FuelTech map.
The idle target is in the calibration, not in a universal rpm. The mixture target is targetlambda, Target Lambda or the configured O2 target. The injector size and rated pressure are in the injector data. The cam target is targetcamangle or VTC Cmd. The VTEC point is in the calibration's VTEC window. The rev limit is in the limiter table. On KPro and FlashPro, the knock threshold is in the knock tables, not in a count from another car.
5. Read fuel against the target, then check capacity
Fuel reading starts with the target. In KManager, compare lambda with targetlambda. In FlashPro, compare Lambda with Target Lambda. In PowerFT, compare the O2 actual with the configured target. A lambda of 0.98 is not lean or rich by itself. It is lean against a 0.90 target and rich against a 1.00 target. The target is the reference, and the reference has to come from the calibration.
More details
A trim is a correction, not a verdict. In KManager, a positive short-term trim means the closed loop is adding fuel at that moment. A lean base table can produce it.
So can low fuel pressure, wrong injector data, or an oxygen sensor that reads a different mixture than the one in the cylinder. A zero trim does not prove the table is right either. The loop may be disabled, limited, stale or working from a sensor that is not valid.
Before you change a fuel cell, overlay mixture, pulse width, duty, voltage, rpm, MAP and loop correction. Add fuel pressure when a characterized sensor exists. Compare pressure across the injector: rail pressure minus manifold pressure, using the same reference for both.
A falling rail gauge reading alone is inconclusive when MAP also falls. If differential pressure falls and the mixture departs from target, investigate delivery before changing the table. Pressure and wideband channels have different acquisition and filter delays; apparent order within those delays cannot establish which event came first.
Sensor delay matters. Exhaust lambda describes earlier combustion, so a throttle step and same-row lambda may belong to different fuel cells. Use stable segments for table learning. A misfire can leave unburned oxygen that produces a lean reading.
Check misfire codes and combustion before treating that reading as a request for more fuel.
A lean row on a closed-throttle overrun can be the fuel cut working. The injectors are off or near off, and the exhaust reads air. Adding fuel to that cell because the trace went lean is the wrong correction. Mark the fuel-cut state first, then exclude it from the fuel-table average.
6. Read the cam, the spark and the knock evidence
The cam pair is a command and an actuator. Compare camangle with targetcamangle in KManager, VTC Actual with VTC Cmd in FlashPro, or the intake phase actual with the configured target in PowerFT. Check the direction, the rate and the settled value at a steady operating point.
- A lag during a transition is normal. A settled error is different. If the actual phase stops short at a steady load, check oil pressure, solenoid operation, sensor zero and mechanical travel. The command may still sit above it. A log cannot measure clearance, and it cannot prove that a 50 degree command is safe for the build. A cam target is a request to the mechanism, not a statement about piston-to-valve clearance.
- The ignition column is the commanded timing in degrees before top dead center. A step in that trace is a change in the command. It is not automatically knock, a knock-retard event or a bad table. Look for the state that changed at the same time. That can be a cam transition, a load step, a temperature compensation, a gear change or a protection limit. Then look at the available knock channels for that calibration.
- Honda knock evidence is not one number. KManager can use the stock knock control tables, and K.Retard is not universally zero. FlashPro exposes Knock Level, Knock Retard and Knock Control where the ECU supports them. A low Knock Control value is not proof that no detonation occurred, and a count is not a calibrated torque loss. A timing notch and a count at the same instant are a prompt to investigate, not a completed diagnosis.
- A log alone cannot establish MBT. The torque plateau needs a torque measurement and a controlled sweep at a fixed state. A log can show that the command moved, that the cam or loop state changed with it, and that a knock channel reacted. That is enough to form a testable cause. It is not enough to declare the timing for the engine.
| Observation | What it can support | What it cannot support |
|---|---|---|
| Lambda above target in a steady closed-loop segment | Measured mixture differs from target; the trim channel shows whether correction is being added | A specific fuel-table cell is lean before pressure, sensor and injector data are checked |
| Lambda above target at fuel cut | Confirmed fuel cut explains the oxygen-rich exhaust reading | A lean mixture during combustion |
| Cam actual below command at a steady load | The phaser is not reaching the request at that point | A failed phaser, a sensor fault or a clearance limit by itself |
| Cam actual below command during a transition | The phaser may still be following the request; inspect its movement | A fault, without the rate and the settled value |
| VTEC output on | The ECU commanded the change | That oil pressure moved the rocker or the high cam is active |
| Spark step with a knock count | The command changed near the same event | That the step was caused by knock, or that the count was real detonation |
| Low Knock Control | The estimator is low at that point | That no detonation occurred |
7. The review sequence, with a stop at each gate
Work through the checks in order. Missing state, targets or required channels limit the conclusion; they cannot justify a table correction.
- Validate the file. Check that relative time rises and the header and units match the software. Check that the sample spacing is plausible and the native log agrees with the CSV at several timestamps. Stop if the time resets, a channel is flat at an impossible value, or a column is blank where the analysis needs it.
- Mark the state of each segment. Separate crank and post-start, warm idle, closed loop, fuel cut, VTEC command, cam transition and steady load. Stop if the state is unknown at the exact segment you want to correct.
- Compare with the target, not with a habit. Lambda against target lambda, cam actual against the command, final spark against the active tables and compensations, duty against the injector data. Stop if the target channel or the injector data is missing.
- Check delivery before the table. Look at pressure, voltage, pulse width, duty and sensor validity. Stop table edits while unexplained differential-pressure or supply-voltage loss remains; account for channel delays before assigning cause order.
- Write one hypothesis and one change. Name the cell or the table, the direction, and the measurement that will show whether it worked. Stop if the change cannot be reversed or cannot be isolated from another change.
- Repeat the same condition. Match load, fuel, temperature and acceleration rate. Check the adjacent cells and the transient on both sides. Keep the failed test as well as the good one.
After a change, check adjacent cells and the transition into and out of the state. Repeat under matched conditions. An improved average can still hide a narrow lean region, so keep both the successful and unsuccessful logs.
8. Two worked cases, labeled as constructed examples
These cases are synthetic. They show the order of the reading, not a repair record from a car.
More details
Case one: the mixture goes lean as delivery falls. In this constructed K24 log, lambda rises from its 0.90 target to 0.98 at steady MAP. At unchanged airflow and with no feedback correction, reaching the target would require 0.98 / 0.90 = 1.0889, or 8.89% more fuel. But the log also shows falling battery voltage and injector differential pressure while pulse width rises. A hypothetical rail gauge reading of 50 psi at 10 psi manifold boost means 40 psi across the injector.
If rail pressure falls to 45 psi with boost still at 10 psi, differential pressure is now 35 psi. These pressures are example readings, not settings. Check pump supply, filter, regulator and sensor scaling before adding fuel.
Repeat after repairing the identified fault and compare pressure and mixture at matched load; the ratio alone does not justify a table edit.
Case two: the mixture moves during a cam transition. A constructed K20 log shows a short lean excursion after the VTEC command while actual phase is still moving. Confirm the lobe command and compare the contributing cam-angle fuel tables, rather than assigning every row to the new commanded angle. Check transport delay and the loop state at the same event. A fuel bump at the crossover can hide an error in the contributing tables. Repeat at the same acceleration rate after the targeted correction; a faster run spends less time in the transition.
No universal idle speed, idle MAP, coolant window, mixture target, injector duty ceiling, cam target, VTEC point, knock count or timing adder. No claim that a log can certify an engine as safe. No claim that one run covers every cell.
Questions
Why does my KManager CSV have no knock retard column?
The documented export names knockcount but not K.Retard or K.Control. Inspect the actual header because exports vary by build. Check native playback for other channels and record the strategy in use. If retard is absent, report it as missing. A missing column cannot establish zero retard.
Is a lean reading at fuel cut a fuel problem?
Not when fuel cut is confirmed. Closed throttle, elevated rpm and zero injector pulse can explain a lean overrun reading. Exclude those samples from mixture averages. For a shift spike, inspect pulse width and any ignition cut as well: unburned oxygen during a misfire or ignition cut can also read lean.
A spike with injection active needs its own diagnosis.
Why does the cam read zero right after a start?
KManager documents a zero, fully retarded intake position for the first 10 seconds after startup. That reading alone does not identify a fault. Evaluate tracking after the lock releases, with a valid command and the required operating conditions met.
Persistent zero actual with a nonzero command needs investigation; it does not identify the failed part by itself.
Why does the actual cam not equal the target in every row?
KManager documents 0.1 to 0.5 seconds of possible lag behind CAMCMD. At 6,000 rpm that spans 10 to 50 crankshaft revolutions. Compare movement and settled error rather than demanding row-by-row equality. The documented lag is context for a transition, not permission to ignore a phaser that never follows.
Does a VTEC command prove the lobe changed?
No. The output proves the ECU requested engagement. Oil pressure, the solenoid and the rocker mechanism still have to respond. Check the command alongside available pressure and airflow or torque evidence. Neither a command bit nor a change in sound directly measures rocker-pin engagement.
Why does 1.80 bar not mean 26 psi of boost?
KManager exports map in bar; a factory Honda MAP sensor measures absolute pressure. With local atmosphere at 1.013 bar, (1.80 − 1.013) × 14.5038 = 11.4 psi gauge. The 26.1 psi figure is absolute pressure. Use measured local atmosphere for the actual car rather than always subtracting sea-level pressure.
Can I compare a KPro log with a FlashPro log row for row?
Compare the same operating state after checking channel definitions, units, sample spacing and filtering. Align events by time rather than row number. Start with a stable segment, then assess transitions. If either file lacks the equivalent channel, describe that limit instead of substituting zero.
Does a low Knock Control value mean no knock?
No. In the documented port-K stock knock strategy, K.Control estimates fuel quality between the 100 and 90 RON references. The scale runs from 0% at the high-octane end to 100% at the low end. A value of 25% nominally represents 97.5 RON.
It is not a fuel tester or an AKI reading. Check whether the file uses percent or a normalized fraction; 0.25 as a fraction means 25%. The estimate changes slowly. Read count, level, retard and operating state together; a low estimate cannot prove that no detonation occurred.
Can one run tune every cell?
No. It visits a limited path through rpm, load, cam angle and lobe state. Record stable coverage where a correction is needed, then verify neighboring regions and transitions. No single gear covers the whole map. Use the actual MAP, throttle and cam traces to establish which region was visited.
What is S.Trim versus L.Trim, and how much is too much?
Short-term trim is the current feedback correction. Long-term trim is learned correction for the applicable operating region. Positive Hondata trim adds fuel; negative removes it. Check the calibration's limits and whether learning and feedback are active. A sustained correction at its limit leaves no further authority in that direction.
There is no single acceptable trim band for every build.
For a constructed steady open-loop example, lambda 0.98 against a 0.90 target gives an 8.89% fuel requirement at unchanged airflow. In closed loop, lambda can already match target while the ECU is adding fuel. Do not then use a zero lambda error to declare the base table correct. FlashProManager's MAP-mode Fuel Change overlay incorporates actual lambda and trims; adding those trims again would count the same correction twice. Check which calculation the export or tool has already performed.
AFCMD and the wideband do not match. Which is the mixture?
AFCMD is the requested mixture; the wideband is an exhaust measurement. Confirm both use lambda or the same AFR display scale, especially on ethanol. Then check sensor warmup, transport delay and loop state. On KPro4 with target lambda tables enabled, changing the target also changes fuel compensation.
Establish the intended target before interpreting the overlay; do not bend the base table to chase an obsolete target.
Why is AFR stuck at 10.7 for the whole log?
A flat reading can be a display limit or an invalid signal. Check sensor/controller errors, heater readiness, selected input and scaling, then compare the ECU reading with the controller's own display if fitted. Repair the signal path and confirm it responds before editing fuel.
The same applies to a reading stuck lean: neither end of the display identifies the mixture when the input is invalid. Pin assignments depend on the ECU and wiring; trace the correct wiring instead of copying a pin swap from another car.
Why do MAP and IAT turn implausible together?
Check the recording and connection before replacing two sensors or changing their scaling. With the engine off and key on, a manifold MAP reading should agree with local atmosphere after converting its reference. After a long soak, IAT should be plausible against ambient.
Reopen the native file and compare another short recording with a known-good direct USB cable. One KPro owner restored communication by removing a USB extension; another reported a loose daughterboard connection that froze the gauges. These are individual repair reports, not proof that every implausible value is a cable fault.
If the problem persists, check ECU supply, grounds and the sensor circuits.
Why is idle timing negative when my ignition table is positive?
The idle controller can retard timing to reduce torque. FlashPro's documented example spans −10° to +15°; it is not a universal limit. Compare actual and target idle speed, air command, voltage and misfire status before editing the main ignition table.
In one KPro idle investigation the owner found a cracked battery terminal after unstable voltage and misfire indications. No completed retest was posted, so the finding does not establish a finished repair.
Why did the ignition value not change by the amount I edited?
Check the active lobe and cam-angle tables, rpm/load interpolation, knock limits and temperature or idle corrections. A neighboring table can still contribute while the cam is between breakpoints. A knock limit can hold timing below the main table without a new count at that timestamp.
Compare the complete calculation before adding the apparent missing degrees back into one cell.
Does a stock calibration with zero knock counts prove the new tune caused knock?
First establish that both calibrations actually produce the count channel. On the older port-K FlashPro strategy, a return-to-stock recording can show zero because stock code lacks the added counting logic. That zero is not a measured baseline.
Compare supported level, control and retard channels with matched fuel and conditions, and investigate the new counts without treating stock zeros as proof.
I could see live FlashPro data. Why is there no saved log?
In FlashProManager laptop logging, F10 starts live viewing; F9 starts or stops recording. Viewing alone does not record the sensor values. Make a short test recording, save the native .fpdl, reopen it and check its duration before the session you need to capture. Onboard logging is a separate recording path. Keep onboard memory until the downloaded native file reopens and contains the event.
What should be in a log I send for review?
Send the native log, untouched CSV and exact calibration revision. Add software/firmware versions, sensor and injector details, fuel blend, gear or load condition and the event time. Include before/after files when testing a change. Remove customer identity and VIN from public examples.
The Tune Report is free. Send the native log, the CSV, the calibration revision and the event note. We return the first checks in order. Honda Full Custom is open at $199 during the launch special ($249 regular).
Every file is compared against our 3 million+ tune values, with a final check-off by our lead tuner.
Tune Report is free if you want this log checked: open the Honda & Acura Tune Report.
A written review of your tune file and one log from your car.
Start a Honda & Acura Tune Report